Under the hood
How we find your buyers.
Reddit has too many posts for anyone to read. LeadFast writes ten searches from your product, runs them, and replaces the ones that stop working. Here is the loop, end to end.
Why ten searches, not one big read
Reddit publishes millions of posts a day. A service that claims to understand every one of them is selling a fantasy: the scan would be too broad to be accurate, and too slow to be early.
LeadFast takes the opposite route. It keeps a small set of searches sharp and reads every post those searches bring back. Two steps, each doing the job it is good at: finding candidates, and understanding them.
The website becomes a product profile
Setup is one URL. We read the website and write down what the product sells, who needs it, and who to skip. Exclusions matter as much as the rest: a post about the wrong audience is noise, even when the words match.
The website is reviewed every day, and corrections override anything the model inferred.
Ten searches, written from the profile
From the profile we write ten complementary searches. Each one is a short phrase or two, aimed at a specific buying problem the product solves. They stay internal: no keyword dashboard, no list to maintain.
Ten different questions cover a problem space better than one long one, and each can be judged on its own results.
The searches run against public Reddit
The searches run across public Reddit posts, not comments, with no Reddit account involved. What comes back is a set of candidates.
Candidates are not a census. Some relevant posts never surface in any search, and we say so instead of implying full coverage.
Every candidate is read and scored
A search brings candidates. It cannot tell whether the person writing is a buyer. So each candidate is read: is there a current problem, is the person looking for a solution, does the product fit, is there urgency?
The read produces a score from 0 to 100. 70 and above is a Strong match, 50 to 69 is Possible. The delivery level sets the bar: every match from 50, or Strong matches only.
Word matching is cheap and wrong often. The read is what keeps an alert worth opening.
The daily exam
Every day, each search is examined against the results it produced.
Searches that stop bringing relevant posts are retired. A search that brings a buyer no other search could find is protected, even while it looks quiet. A retired search is only replaced when a candidate replacement has been validated against the product: nothing leaves on a hunch.
The replacements come from the same theme, meaning the product problems, phrased another way. Over time the ten survivors are the ones producing the most relevant posts for this product.
That is the improvement: not a smarter model every week, but a set of searches that keeps earning its place.
Delivery, in under a minute
Each match arrives with the post, the reason it matched, and a short Suggested Reply. Slack, Telegram, email, or webhook.
Post to alert in under one minute, measured from publication to destination acceptance. Actual delivery times can vary. Sent means at least one destination accepted the alert. Everything else stays visible as not sent.
What the loop shows, and its limits
The dashboard reports posts analyzed and matches delivered over the last 30 days. The daily review uses those observed results to keep, retire, or replace searches.
There is no census, no ranking, and no promise of leads. A post that never appears in the searches is a post the model never reads.
What we will never do
- We will never post, comment, vote, or message on Reddit.
- We monitor public posts, not comments.
- We do not promise coverage, ranking, or leads.
- A post that never shows up in the searches is a post the model never reads.